We sought to understand the genesis of the t(9;22) by characterizing genomic breakpoints in chronic myeloid leukemia (CML) and BCR-ABL-positive acute lymphoblastic leukemia (ALL). BCR-ABL breakpoints were identified in p190 ALL (n ¼ 25), p210 ALL (n ¼ 25) and p210 CML (n ¼ 32); reciprocal breakpoints were identified in 54 cases. No evidence for significant clustering and no association with sequence motifs was found except for a breakpoint deficit in repeat regions within BCR for p210 cases. Comparison of reciprocal breakpoints, however, showed differences in the patterns of deletion/insertions between p190 and p210. To explore the possibility that recombinase-activating gene (RAG) activity might be involved in ALL, we performed extra-chromosomal recombination assays for cases with breakpoints close to potential cryptic recombination signal sequence (cRSS) sites. Of 13 ALL cases tested, 1/10 with p190 and 1/3 with p210 precisely recapitulated the forward BCR-ABL breakpoint and 1/10 with p190 precisely recapitulated the reciprocal breakpoint. In contrast, neither of the p210 CMLs tested showed functional cRSSs. Thus, although the t(9;22) does not arise from aberrant variable (V), joining (J) and diversity (D) (V(D)J) recombination, our data suggest that in a subset of ALL cases RAG might create one of the initiating double-strand breaks.